How to Migrate from ChatGPT to Claude (Step by Step)

To migrate from ChatGPT to Claude, export your ChatGPT data for safekeeping, import your accumulated context through Claude's memory import, then rebuild the rest of your workflow. Claude can now carry over what another AI knows about you, but your chat history and custom GPTs do not transfer — you recreate those as Projects.

Claude memory import and plan availability verified 31 July 2026 against Anthropic's help center.

That is the honest version of the answer, and it saves you a lot of frustration up front. Plenty of people search for a magic button that sucks their entire ChatGPT life into Claude in one click. It does not exist, and any tool promising it is not official. What actually works is a short, deliberate migration: move your data so you keep a record, then rebuild your workflow so Claude fits how you actually work. This guide walks through both, step by step, so you can switch from ChatGPT to Claude without losing anything that matters. This reflects real hands-on time with the feature, not just its documentation.

Claude currently spans models like Opus 5, Sonnet 5, and Haiku 4.5, and this guide is reviewed regularly so the steps stay accurate — the export itself kicks off as soon as you confirm it.

If you are still weighing the decision itself, our breakdown of whether Claude is better than ChatGPT is a good companion read, and if the question is really about which subscription to pay for, see Claude Pro vs ChatGPT Plus. This article assumes you have already decided and just want the cleanest path across.

Why there is no one-click way to migrate from ChatGPT to Claude

Different companies, different data formats, and no shared import API. You copy what is portable — prompts, reference documents, project structure — and consciously recreate what is not, which usually sharpens your setup rather than degrading it.

ChatGPT and Claude are built by different companies (OpenAI and Anthropic), store data in different formats, and expose no shared import API. So when you migrate from ChatGPT to Claude, you are not "porting an account" — you are copying the parts that are portable and consciously recreating the parts that are not.

The good news: most of what makes you productive is portable. Your prompts, your reference documents, your working style, and your project structure all move over with a little manual effort. The thing that does not move is the invisible state ChatGPT has accumulated about you — your chat history, its memory of your preferences, and any custom GPTs you built. You recreate those, and honestly, most people find the rebuild sharpens their setup rather than degrading it.

Step 1 — Export your ChatGPT data first

In ChatGPT open Settings, then Data controls, then Export data. OpenAI emails a link to a zip holding conversations.json and an HTML viewer. Skim it for reusable prompts and reference material; that list drives everything after.

Before you change anything, grab a backup. Even if you never open it again, having a record means you can migrate from ChatGPT to Claude at your own pace instead of racing the clock.

In ChatGPT, open Settings → Data controls → Export data, then confirm the export. OpenAI emails you a download link (it can take a few minutes to a few hours) pointing to a .zip file. Inside you will find a conversations.json file plus an HTML viewer that renders your whole chat history in the browser. That archive is your safety net.

Open the HTML file and skim it. You are looking for three things: prompts you reuse constantly, reference material you paste often (style guides, product specs, boilerplate), and any clever multi-step instructions worth keeping. Copy those into a scratch document — that becomes your migration shopping list. Everything else in the export is just a historical record you can keep on disk.

Step 2 — Know what does not transfer automatically

Chat history, custom GPTs, and GPT Store items do not come across, so you rebuild them. Memory is the exception: Claude ships an official import that turns a plain-text dump of what another assistant knows about you into Claude memory entries.

This is where honest expectations save you time. When you switch from ChatGPT to Claude, the following do not come across on their own — you recreate each one:

• Chat history. Your ChatGPT conversations stay in ChatGPT (and in your export). Claude starts with a clean slate.
• Custom GPTs. The custom GPTs you built or installed do not exist in Claude. You rebuild the useful ones as Claude Projects.
• Memory. This one can follow you, and it's the part of the job that got easier. Claude ships a memory import that takes a plain-text dump of what another assistant knows about you and turns it into Claude's own memory entries — see our walkthrough of Claude import memory. It's available on Free, Pro, Max and Team, so you can migrate from ChatGPT to Claude and bring your context across before you pay for anything.
• GPT Store items. Third-party GPTs have no direct equivalent; you replace them with Claude skills, connectors, or a well-configured Project.

None of this is a dealbreaker — it is a one-time setup cost, usually an hour or two spread across your first week.

Step 3 — Map every ChatGPT feature to its Claude equivalent

Custom GPTs become Projects, custom instructions become styles, the GPT Store becomes skills and connectors, browsing becomes web search, and data analysis becomes file uploads. Work the table row by row against your shopping list.

Here is the core of the migration. Almost every ChatGPT feature has a Claude counterpart; the names differ. Use this table as your translation key when you migrate from ChatGPT to Claude.

Work through the table row by row against your migration shopping list from Step 1. For each ChatGPT habit you had, decide which Claude home it belongs in. A recurring "write my weekly newsletter" GPT becomes a Newsletter Project. Your global "always be concise, use British English" instruction becomes a custom style. Reference PDFs you kept re-uploading become Project files you upload once.

Step 4 — Bring your best prompts over

Prompts are the most portable thing you own, and good prompt engineering carries across largely unchanged. Expect small differences, since Claude follows explicit structure and clear role framing closely. Test your top five side by side first.

Good news: prompts are the most portable thing you own. The prompt engineering you did in ChatGPT mostly works in Claude with little or no change. Copy your strongest prompts straight across and test them.

You will notice a few style differences. Claude tends to follow structure and explicit instructions closely, responds well to clear role framing, and is comfortable with longer, document-heavy context thanks to its large context window. If a prompt relied on ChatGPT-specific quirks, tidy it up: state the goal plainly, give an example of the output you want, and specify format. If you want to go deeper, our guide to how to use Claude AI covers the everyday basics of getting great results.

A practical tip: do not migrate prompts blindly. Run your top five against a real task in both tools side by side for a day. You will quickly learn where Claude needs a slightly different nudge, and you will trust the results more.

Step 5 — Set up Projects and custom styles for recurring work

Projects bundle instructions, reference files, and related chats around one recurring job, which is the closest replacement for a custom GPT. Pair them with custom styles so tone stays consistent. Do not shortcut this step.

If you do the same kind of work repeatedly — client emails, code reviews, research summaries, marketing copy — Projects are where Claude pays you back for migrating. A Project bundles instructions, reference files, and conversation history around a single job, so you stop re-explaining context every session.

Create one Project per recurring workflow. Add a clear instruction block ("You are my editor for a fintech newsletter; house style is plain, skeptical, no hype"), upload the reference documents that job always needs, and let the Project accumulate relevant chats. This is the closest thing to rebuilding a custom GPT, and most people end up preferring it because the reference files live inside the workspace.

Pair Projects with custom styles so your tone stays consistent across everything. Define a style once — formal, casual, technical — and apply it anywhere. Together, Projects and styles replicate the "trained assistant" feeling you had in ChatGPT, minus the black-box behavior.

From running Claude in our own daily work, this is the step we'd push you not to shortcut. We long ago moved our recurring context out of any assistant's invisible memory and into explicit, checked-in instruction files, and the payoff showed up exactly here: when you keep your standing context written down instead of trusting a hidden memory, the "nothing transfers" worry basically evaporates, because you can paste it into a fresh tool in minutes. That same habit is why, whenever we move context between tools, the job that people dread when they migrate from ChatGPT to Claude feels like a copy-paste rather than a rebuild.

Step 6 — Give Claude a fair trial (and keep your ChatGPT export)

Give Claude two weeks of real work as your default before judging it, not a single pasted prompt. Keep your ChatGPT account and export throughout — Claude exports your chats too, so portability runs in both directions.

The most common mistake is a half-hearted switch: people open Claude, paste one prompt, get a different-flavored answer, and give up. Migration is a habit change, not a single test. Commit to using Claude as your default for two weeks of real work before you judge it.

During the trial, keep your ChatGPT account and export around. There is no rush to delete anything. If you later decide Claude is home, you can archive the export. And because Claude lets you export your Claude chats too, you are never locked in — the same portability that made this migration possible works in both directions.

For the full picture of what Claude can do out of the box — vision, artifacts, web search, connectors — the official Claude overview is worth a skim, and Anthropic's help center covers account, billing, and data questions if anything comes up mid-switch.

How long does it take to migrate from ChatGPT to Claude?

Requesting the export takes minutes and the rebuild takes roughly one to two hours, comfortably spread across a week: back up on day one, map features and prompts by day three, build Projects by day five, judge in week two.

You do not need to do all six steps in one sitting. Here is how a smooth switch usually unfolds.

Spread across a week, the whole thing is light. The export takes minutes; the rebuild is where the value is, and it doubles as a spring-clean of prompts and reference files you had let go stale in ChatGPT.